Parallel Debugging

Framework for debugging complex issues using the Analysis of Competing Hypotheses (ACH) methodology with parallel agent investigation.

When to Use This Skill

• Bug has multiple plausible root causes
• Initial debugging attempts haven’t identified the issue
• Issue spans multiple modules or components
• Need systematic root cause analysis with evidence
• Want to avoid confirmation bias in debugging

Hypothesis Generation Framework

Generate hypotheses across 6 failure mode categories:

1. Logic Error

• Incorrect conditional logic (wrong operator, missing case)
• Off-by-one errors in loops or array access
• Missing edge case handling
• Incorrect algorithm implementation

2. Data Issue

• Invalid or unexpected input data
• Type mismatch or coercion error
• Null/undefined/None where value expected
• Encoding or serialization problem
• Data truncation or overflow

3. State Problem

• Race condition between concurrent operations
• Stale cache returning outdated data
• Incorrect initialization or default values
• Unintended mutation of shared state
• State machine transition error

4. Integration Failure

• API contract violation (request/response mismatch)
• Version incompatibility between components
• Configuration mismatch between environments
• Missing or incorrect environment variables
• Network timeout or connection failure

5. Resource Issue

• Memory leak causing gradual degradation
• Connection pool exhaustion
• File descriptor or handle leak
• Disk space or quota exceeded
• CPU saturation from inefficient processing

6. Environment

• Missing runtime dependency
• Wrong library or framework version
• Platform-specific behavior difference
• Permission or access control issue
• Timezone or locale-related behavior

Evidence Collection Standards

What Constitutes Evidence

Citation Format

Always cite evidence with file:line references:

**Evidence**: The validation function at `src/validators/user.ts:87`
does not check for empty strings, only null/undefined. This allows
empty email addresses to pass validation.

Confidence Levels

Result Arbitration Protocol

After all investigators report:

Step 1: Categorize Results

• Confirmed: High confidence, strong evidence, clear causal chain
• Plausible: Medium confidence, some evidence, reasonable causal chain
• Falsified: Evidence contradicts the hypothesis
• Inconclusive: Insufficient evidence to confirm or falsify

Step 2: Compare Confirmed Hypotheses

If multiple hypotheses are confirmed, rank by:

1. Confidence level
2. Number of supporting evidence pieces
3. Strength of causal chain
4. Absence of contradicting evidence

Step 3: Determine Root Cause

• If one hypothesis clearly dominates: declare as root cause
• If multiple hypotheses are equally likely: may be compound issue (multiple contributing causes)
• If no hypotheses confirmed: generate new hypotheses based on evidence gathered

Step 4: Validate Fix

Before declaring the bug fixed:

• Not done: Fix addresses the identified root cause
• Not done: Fix doesn’t introduce new issues
• Not done: Original reproduction case no longer fails
• Not done: Related edge cases are covered
• Not done: Relevant tests are added or updated